Home CTV Roundup Swivel Wants To Create Agentic Infrastructure For CTV

Swivel Wants To Create Agentic Infrastructure For CTV

SHARE:

Agentic media buying is the No. 1 topic du jour in ad tech discussion circles. But how does agentic buying even work when it comes to buying TV and streaming ads? Those channels remain more old school and anchored by direct human relationships, at least compared with web display advertising.

TV is a media channel that’s going to have its own particular setup and adoption curve.

At least that’s the thesis behind Swivel, one such agentic ad-buying startup. 

Swivel is a publisher tech startup founded by ad tech entrepreneurs and investors: the former FreeWheeler Frans Vermeulen and SpringServe co-founders Matt Dearborn, Rich Lin and Joseph Hirsch. The startup aims to make a name for itself as a company that can automate the manual tasks in publisher ad operations, from campaign optimization to pricing adjustments. 

In other words, Swivel wants to be an “end-to-end system” of agents that handles all aspects of campaign management, from media planning to optimization and reporting, Hirsch told Streamlined. The eventual goal, he said, is for an agentic infrastructure that spans the CTV advertising chain in particular and is part of media sellers’ yield optimization.

AI Ad-option

AI product adoption may seem inevitable. But Hirsch acknowledged that there is no cakewalk in developing agentic solutions for every step in the CTV chain and to gain wide adoption. 

Initially, Swivel was “very active in the middle to tail end of the agentic media workflow,” Hirsch said, especially where yield optimization is concerned. But now, he said, “we’re much more active at the front end of the agentic workflow,” referring to campaign functions before any media is purchased, such as planning, forecasting and ad ops rule-setting for guaranteed deals. 

Rather than a single consolidated product, Swivel has multiple agents each responsible for parts of a campaign buy, such as budgeting, bid trafficking, pacing, yield management and post-campaign reporting, just to name a few, Hirsch said. 

Which is certainly not to say that all publishers are comfortable entrusting their entire media-selling process to AI. Many publishers are interested in using one or two agents, “so as not to disrupt their entire ad ops workflow,” he said. It’s common for clients to pick and choose what they would prefer to outsource to the AI model.

Plus, agentic automation isn’t necessarily simplification. There is a real concern that agentic platforms and MCPs just become another tangle in the fragmentation of CTV media. Plus, with every vendor evolving into an AI solution of some sort, CTV advertisers and ad sellers might easily feel like there are too many agents in the kitchen. (Had to.) 

Hirsch said this type of fragmentation is addressed in part by specifications within the Ad Context Protocol (AdCP), a set of standardized protocols for online ad supply chain agents to communicate and negotiate. AdCP includes a specification called adagents.json that allows publishers to declare which agents are authorized to sell their inventory, not unlike sellers.json, Hirsch said. (Editor’s note: The AdCP is collaborative with Prebid and somewhat in opposition to the IAB Tech Lab’s AAMP and ARTF frameworks.)

The case for agentic CTV

Agentic media buying on TV screens faces a long, uphill climb, even with optimistic expectations. But there are real perks to incentivize adoption by CTV advertisers and sellers. And those benefits are distinct even from programmatic. 

For example, certain non-standard CTV ad formats – such as sponsorship placements, home screen ads and pause ads – can differ across streaming TV apps, smart TV models and CTV buying channels. And without ad standardization across publishers, these CTV buys are difficult to set up programmatically, Hirsch said. 

Ad tech vendors and industry orgs have recently announced standardized ways to integrate home screen and pause ads. But it takes time to build wider buy-in.

With agentic media buying, on the other hand, it’s possible to automate ad sales for formats that are unavailable or difficult to buy programmatically, Hirsch said. Along with nontraditional ad formats, he said, sequential media (having creatives appear in a certain order within a program) is another type of buying strategy that can be managed agentically but not programmatically.

So, what’s next for Swivel? 

Today, “our focus is still 100% CTV,” Hirsch said, but the startup intends to expand into the retail media space, which is especially promising because of the rising overlap between retail media and CTV, he said. Just look at Walmart, Vizio and Vibe.co.

Must Read

AppLovin Is Back To Smashing Earnings, But Investors Are Slow To Buy

AppLovin is growing at a blistering pace, but Wall Street investors remain cautious about the business.

Gamers Love Ad Blockers. So This Site Found Another Way To Show Them Ads

High-end ad blockers mimic some of the same behavior as bot traffic. So WiseGuyGaming repurposed the tools it was already using to identify bots to reveal its ad-blocking audience as well.

Criteo Is A Survivor, But Expanding Beyond Retail Media Will Be Its Toughest Challenge Yet

Criteo has weathered considerable storms in its day, including a painful moment in March 2020 when its market cap bottomed out below $500 million. Can it pull off yet another turnaround?  

Privacy! Commerce! Connected TV! Read all about it. Subscribe to AdExchanger Newsletters

Sports Leagues Can’t Stop Putting Ad Revenue Ahead Of Fan And Player Interests

Disaffection and outright anger toward US sports institutions run deeper than logo-cluttered fields. Leagues and teams are sacrificing player safety and fan enjoyment for marketing dollars.

Comic: Traffic Jam

With Search Traffic Down 70%, People Inc. Bets On Revenue Beyond Site Visits

People Inc.’s referral traffic from Google search is down 70% over the past two years, and by 26% in Q3 alone. But its digital revenue is still growing, albeit slowly.

Chord Buys Its Rival Zenlytic To Build An Agentic Version Of A CDP

On Tuesday, Chord, a six-year-old commerce tech startup, acquired Zenlytic, an AI startup with an anthropomorphized data analyst agent named “Zoë,” for $5 million.